DEVELOPMENT OF FIELDBUS ARCHITECTURE FOR TELEOPERATION AND DISTRIBUTED CONTROL OF A MOBILE ROBOT

Transcription

1 DEVELOPMENT OF FIELDBUS ARCHITECTURE FOR TELEOPERATION AND DISTRIBUTED CONTROL OF A MOBILE ROBOT Eduardo Paciência Godoy, epgodoy@yahoo.com.br Arthur José Vieira Porto, ajvporto@sc.usp.br Department of Mechanical Engineering, EESC - Engineering School of São Carlos, USP University of São Paulo Avenida Trabalhador São carlense, 400 CEP , São Carlos, São Paulo, Brazil Ricardo Yassushi Inamasu, ricardo@cnpdia.embrapa.br EMBRAPA - Brazilian Agricultural Instrumentation Research Corporation Rua XV de Novembro, 1452 CEP , São Carlos, São Paulo, Brazil Abstract. A current trend in the agricultural area is the development of mobile robots and autonomous vehicles for in-field data acquisition and remote sensing. These robots and vehicles developed with the same technologies existing in agricultural machinery can be more efficient doing these specific tasks than traditional large tractors. One of the major challenges in the design of these robots is the development of the electronic architecture for the integration and control of the several devices related to the motion, navigation, data acquisition and communication systems. An electronic architecture must be robust and reliable, provide quick and ease maintenance and have modularity and flexibility to allow future expansions and connections of new equipments. Recent applications of mobile robots have used distributed architectures based on fieldbus networks to meet these requirements. This work describes our approach to designing and implementing a teleoperated distributed control system based on CAN protocol for a mobile agricultural robot. The discussions are focused on the developed electronic fieldbus architecture, the wireless communication system for teleoperation and the distributed control over the CAN network.. The evaluation of the developed system was based on the analysis of performance parameters such as motors response and architecture time delay obtained with the robot operation. The results show that the developed fieldbus architecture can be applied for teleoperation and distributed control of agricultural mobile robots meeting the requirements for an accurate robot movement and an acceptable response time for control commands and supervision. Keywords: agricultural robot, CAN protocol, networked control design, teleoperation 1. INTRODUCTION An increase in the application of the automation and informatics in the agricultural area can be observed in recent times. New agricultural practices, related to the Precision Agriculture, have enhanced the importance in the research of embedded sensors and communication networks (Auernhammer and Speckman, 2006) for the study of spatial variability and remote sensing applications. New technologies and devices for real-time data acquisition and actuation have been released to equip agricultural machineries to support these practices and automated them (Auernhammer, 2004). A strong tendency is development of mobile robots and autonomous vehicles for application in specific tasks, improving the efficiency and giving better results (soil compactation reduction and machine operator absence) when compared with the use of traditional large tractors and implements(blackmore and Griepentrog, 2006). Autonomous vehicles and mobile robots have been widely used in industrial production and warehouses, where a controlled environment can be guaranteed. In agriculture areas, research into driverless vehicles has always been a dream but serious researches started in the early 1960 s (Blackmore et al., 2005). In recent years, the development of these vehicles has experienced increased interest. This development has led many researchers to start developing more rational and adaptable vehicles. These vehicles should be capable of working 24 hours a day all year round, in most weather conditions and have the intelligence embedded within them to behave sensibly in a semi-natural environment over long periods of time, unattended, while carrying out a useful task (Pedersen et al., 2005). In scientific literature can be find studies that seek to adapt business agricultural machinery to make agricultural platforms (autonomous vehicles or mobile robots) as can be seen in Reid et al. (2000) and Keicher and Seufert (2000). A more recent trend is the development of platforms built specifically for agricultural autonomous vehicles or robots as can be seen in Åstrand and Baerveldt (2002), and Bak and Jakobsen (2004). However, the development of these platforms presents two challenges (Blackmore et al., 2004): developing a physical structure suitable for the agricultural environment, and develop an architecture to integrate the various electronic devices allowing future expansions through the addition of new devices. A technology that has strong potential to be applied on these devices interconnection is the distributed control systems over communication networks or fieldbus. Fieldbus based control systems have replaced the traditional centralized control systems because of several benefits such as reduced cost and amount of wiring, increased reliability and interoperability, improved capacity for system reconfiguration and ease of maintenance (Moyne and Tilbury, 2007). Although the fieldbus distributed control systems

2 offers several advantages over traditional centralized control systems, the existence of communication networks make the design and implementation of these solutions more complex. Networked control systems impose additional problems inherent in control applications that are usually difficult to meet due to the variations and uncertainties introduced by the fieldbus: delays, jitter, bandwidth limitations and packet losses (Baillieul and Antsaklis, 2007). In the agriculture area, the chosen of the CAN protocol (Bosch, 2006) as communication network due to its low cost of development and large acceptance and success for embedded electronics in the automotive area. The use of CAN in the agricultural area is confirmed in Suvinen and Saarilahti (2006) and its application to autonomous vehicles and mobile robots is presented in Nagasaka et al., (2004) and Darr; Stombaugh and Shearer (2005). The implementation of the ISO11783 standard represents the standardization of the CAN protocol to the agricultural area and constitutes the main target of development as described in Benneweiss (2005). Following this guideline, this paper describes the design and implementation of a teleoperated distributed control system based on CAN protocol for a mobile agricultural robot. The Wireless Ethernet to fieldbus architecture is detailed presented and the distributed robot control over the CAN network is designed and discussed. Performance parameters such as motors response and architecture time delay obtained with the robot operation allows verify that the developed teleoperated fieldbus architecture can be applied to distributed control of agricultural mobile robots. 2. THE MOBILE ROBOT ARCHITECTURE 2.1 Robot Structure and CAN Fieldbus System The agricultural mobile robot was designed to be used as an experimental platform for development of control, navigation and data acquisition technologies to the agricultural area. The major application of the robot is to do the remote sensing of agronomic parameters of most important Brazilian culture in large areas. It doesn t require actions that demand high power, as in agricultural operations, but only moving efficiently in this environment. The mechanical structure, showed in Fig. 1, was designed by the studying of work conditions required in field and desired characteristics of the project. It was established that the structure should be in portico with 2m of height and 2,5m of length, capable of operating in cultures up to 1.5 m of height, with adjustable gauge (width of 1,5 to 2,5m) to operate in various row spacing cultivation. To accomplish this, the system was designed in independent modules (side frame number 1 and top frame number 8 in Fig. 1), together by telescopic bars (number 10), to meet the maximum possible situations. To accomplish this, the system was designed in independent modules (side frame number 1 and top frame number 8 in Fig. 1), together by telescopic bars (number 10), to meet the maximum possible situations. The steering module (number 6), the propulsion module (number 5) and central box (number 9) complete the system. The structure also should be light and flexible compared with commercial agricultural vehicles, with the possibility to insert new sensors and actuators. The side boxes (number 7) contain the electronic systems to communicate with the CAN fieldbus and the motor controllers and also protect these devices from weather, dusty and vibrations. It is important to observe that heavier items in the robot such as batteries (number 4), propulsion and steering systems and side boxes are at least one meter of the soil, contributing to lower structure center of gravity, increasing its stability on sloping land. The robot architecture with distributed CAN fieldbus was designed symmetrically between right and left sides of the structure, which allows the homogenous distribution of weight, simplifies the development, reduces design time and costs and the amount of cables, and eases the maintenance of equipments installed in the system. Figure 1. Architecture of the Mobile Robot with the CAN Fieldbus

3 The propulsion system of the robot needs to have accuracy in direction, low power consumption and low cost. Propulsion systems with wheels are cheap and, in function of the low need for traction and load to be distributed, meet the needs of this project. In this project, we adopted a four wheels system (number 3 in Fig. 1) and to increase the ability of vehicle pull in adverse conditions, we adopted independent traction in each wheel. Each propulsion system is composed by a Roboteq AX2850 controller, a Bosch GPA 750W 24V DC motor, a 75:1 reduction system (25:1 of a planetary gearhead plus a 3:1 crown, pinion, chain transmission) and a Hohner Serie 75 incremental encoder with 100 pulses per revolution. Among the steering systems found, there are differential steering, articulated steering and wheel steering. The wheel steering or Ackerman system is the most used by road vehicles for steering. This methodology describes the relation between the angles of outside and inside the wheel in a turn. In function of structure configuration is in portico format and with adjustable gauge, it was chosen a system that could be independent for each wheel, with easy construction and accuracy of steering, so we opted by the system Ackerman in front wheels. Each steering system is composed by a Maxon Motor kit (EPOS 70\10 positioning controller with CAN interface, a RE40 150W 48V DC motor with a 230:1 reduction planetary gearhead GP22C and MR incremental encoder with 500 pulses per revolution). For integration (communication by the network, information exchange and control) between electronic devices, it was deployed a CAN fieldbus network based on ISO11783 protocol in the agricultural mobile robot. An electronic control unit (ECU), or CAN interface, develop in our laboratory (Sousa, 2002) was used for this devices integration. This ECU is mainly composed by a microcontroller with a CAN controller (PIC18F258), a CAN transceiver (MCP2551) and a RS232 transceiver (MAX232). The microcontroller provides the logic operations for the CAN protocol communication and I/O data acquisition. The transceivers provides switching between the digital TTL logic of the microcontroller and analog signals required on the CAN bus and the RS232 port. The CAN network developed uses a 250Kbits/s transmission speed and not only enables the integration of sensors, actuators and computer systems relative with tasks of navigation (motor controllers, GPS and digital compass), but also enables the devices integration related to data acquisition of agronomical variables, which will eventually compose the architecture of the robot. In the architecture developed, the mobile robot is teleoperated. A teleoperation station showed in Fig. 1, has the function of managing the operations performed by the robot, permitting planning, controlling and monitoring tasks in real-time via a Wireless Ethernet network based on IEEE for data communication performed through a VNC connection. A directional antenna in both systems (teleoperation station and robot) allows the communication and teleoperation up to 5 km of distance. The industrial computer in the mobile robot provides the connection between the teleoperation commands and the devices in the CAN network. As the industrial computer uses a traditional operational system (Windows XP), the architecture does not guarantee real-time determinism. In order to obtain a real-time predictability for the system, a real-time operational system is required. Aroca (2008) presents a detailed study about real-time operating systems (RTOS) that can be used in digital control systems for automation and robotics applications. This work can be used as guide for the definition and implementation of a RTOS for the mobile robot presented. 3. DISTRIBUTED CONTROL OF THE ROBOT 3.1. Design of the CAN-Based Control System As described in Johansson, Torngren and Nielsen (2005), in CAN-based networks data are transmitted and received using message frames that carry data from a transmitting node to one or more receiving nodes. An identifier, unique throughout the network, labels each message of the node and its value defines the priority of the message to access the network. The CAN protocol is optimized for short messages and uses a CSMA/CD with NDBA (Carrier Sense Multiple Access / Collision Detection with Non-Destructive Bitwise Arbitration) arbitration access method. The bit stream of a transmission is synchronized on the start bit, and the arbitration is performed on the following message identifier, in which a logic zero is dominant over a logic one. Figure 2 presents the flowchart of the distributed control in the mobile robot. According to the flowchart, the user cans teleoperated the robot by selecting between two control methods. Sending manual commands or setting predefined commands to control the mobile robot. Computer of Teleoperation Station Manual Command Predefined Commands or Trajectory Define Control Method Command Scheme Supervision and Control Wireless Wireless Mobile Robot Industrial Computer Control Inputs CAN Network CAN Network Motors PID Controllers Motors Feedback, Data of Navigation Devices and Sensors Figure 2. Control Flowchart of the Mobile Robot

4 The predefined commands do not allow autonomous navigation only defined trajectories (for example walk in straight-line, do curves of user defined degrees) that simplify the necessary user commands to be sent to the robot. All commands defined by the user are transmitted to the mobile robot via a wireless digital link based on IEEE standard performed by a VNC connection. The industrial computer in the mobile robot functions like a gateway. All information received via wireless link is transmitted into messages in the CAN network and vice-versa. The industrial computer is responsible too for the vision acquisition of the camera in the robot. The defined commands received via wireless are translated in control inputs for the motor controllers of the mobile robot. These control inputs are transmitted into messages in the CAN network. The ECUs in the robot receive the CAN messages with the control inputs and act in the motor controllers. The motor controllers use discrete-time PID controllers to control the robot motion. The motors feedback (encoder information about traction speed and steering position) and the data from other devices and sensors connected in the robot are transmitted too via the CAN bus to the industrial computer. And this information is sent back via wireless to the computer in the teleoperation station and is presented in the supervision and control software. Using the information about the robot, the user can decide how to actuate and control the robot movement, finishing the flowchart of the robot distributed control. Experiments were done to design and define the PID controller gains for the motors control (speed control for propulsion motors and position control for guidance motors) of the mobile robot. The PID controllers for speed control uses a sampling time of 16ms and were defined with the same gains (P=2, I=1.5 and D=1) because the four propulsion systems have the same equipment. The PID controllers for position control used a sampling time of 1ms. The values defined for the controller gains were P=275, I=20 and D=200. The gains of the PID controllers are defined to achieve a suitable operation of the robot movement. The PID controllers for the traction motors need a soft start (speed increases like a ramp) because the high current demanded for the motors startup (the DC motors used in the robot propulsion have 750W and can demand up to 100A in startup if no soft startup is used). The PID controllers for the guidance motors need to present no overshoot and low movement speed. The required maximum value for the robot guidance position change is equal to 15 rpm to not influence in the robot guidance control Robot Operation and Performance Analysis With the teleoperated architecture implemented in the mobile robot and the controllers designed to the CAN network, the mobile robot operation could be evaluated. We performed field tests using the mobile robot to evaluate the teleoperated fieldbus architecture (response time of user commands and quality of motors control). In experiments, the user controls the mobile robot navigation and the feedback information is analyzed to check the operability and accuracy of the robot movement. The architecture time delay (response time of user commands transmission and feedback information for supervision) is also evaluated to verify its possible influence on the robot operation and supervision. Figure 3 presents the graphic of the step responses of the tractions motors of the robot for speed control. According to the propulsion design for the mobile robot, one rpm is equal to 0,035 m/s and its the maximum velocity achieved is equal to 1,5 m/s. (a) (b) Figure 3. Test for Mobile Robot Speed Control: (a) Detail of the Step Response showed in the Rectangle a in Graphic (b) (red is the setpoint required and blue is the robot velocity), (b) Step Responses of the Traction Motors Figure 3 shows the accuracy of the traction motors response related to the setpoint required. No steady state error can be found and the motors velocity increased like a ramp indicating the correct design of the PID controllers for mobile robot traction. All the four traction motor present the same response because the CAN network allow the transmission of messages in a diffusion mode. The industrial computer of the robot needs to transmit just one message in the CAN network and all ECU can receive this message and control the motors at the same time. This strategy do that the distributed control system do not present time delay or lack of synchronism in control commands reception.

5 According to the feedback information obtained with the mobile robot traction test, the highest current used by the traction motors was 30A in maximum speed. In normal operation, between 0,5 to 1 m/s, the measured current was up to 18A. With theses values we can estimate operation autonomy for the robot of 4 hours with the batteries currently used. The temperature of the propulsion controllers was also measured to verify possible problems. In the tests done, the temperature does not exceeded 50 C that is less than the maximum value (80 C) acceptable for the controller. Figure 4 presents the graphic of the step responses of the guidance motors of the robot for position control. a) (b) Figure 4. Test for Mobile Robot Guidance Control: (a) Detail of the Step Response showed in the Rectangle a in Graphic (b) (red is the setpoint required and blue is the robot guidance position), (b) Step Responses of the Guidance Motors Figure 4 shows the accuracy of the guidance motors response related to the setpoint required. No steady state error and overshoot can be found and an appropriate low speed was achieved for the guidance movement. According to the Fig. 4(a), a guidance position change of 90 degrees takes approximately one second (1s) to occur resulting in a movement speed less than the 15 rpm defined in the guidance design for a suitable robot navigation. These results indicate the correct design of the PID controllers for mobile robot guidance. The CAN bus load was measured in the robot operation and was less than 10%. This value is low and shows that the CAN-based distributed control of the mobile robot has capabilities to future expansions or new devices connections and increase in the data load and message traffic. The architecture time delay was evaluated by measuring the response time of user commands since beginning of the command transmission by the teleoperation station until the distributed control over the CAN network. An example of this response time can be seen in the start of the Fig. 4(a) that shows a response time of approximately 40 ms. The values measured do not exceed 100ms what indicates that the architecture time delay does not affect the robot operation and supervision and is adequate for teleoperation of the agricultural mobile robot developed. Finally the results of the tests showed that the agricultural mobile robot could be teleoperated. Even though it is desirable to improve the mobile robot control, the present level of accuracy and the architecture time delay is sufficient for teleoperation and remote sensing, and the results indicate that the developed architecture and the distributed control are useful for agricultural mobile robot operations. Future work will be done to improve the mobile robot control and to develop autonomous navigation capabilities to the mobile robot. Also the use of a real-time operational system will be evaluated to provide predictability and guaranteed determinism for the robot architecture. 4. CONCLUSIONS This paper presented the development of fieldbus architecture for teleoperation and distributed control of an agricultural mobile robot. The application of the CAN protocol as the fieldbus provided an efficient platform to develop the distributed control system of the robot. Individual control nodes or electronic control units (ECUs) reduced the computational load of the task computer by implementing feedback control logic at the ECUs and ease the data communication between the devices of the robot. The CAN network allowed the transmission of messages in a diffusion mode doing that the distributed control system do not presented time delay or lack of synchronism in control commands reception (for example between the four controllers of the propulsion motors). Tests were performed using the mobile robot to evaluate the fieldbus architecture develop in terms of the teleoperation system and the distributed control over the CAN network. The values measured for the response time of user commands do not exceed 100ms what indicates that the architecture time delay does not affect the operability of the robot and supervision and is sufficient for teleoperation of the agricultural mobile robot developed. The PID controllers used to control the propulsion and guidance motors were designed to achieve a suitable and precise operation for the mobile robot navigation.

6 The results of the tests demonstrated that the developed fieldbus architecture can be applied for teleoperation and distributed control of agricultural mobile robots meeting the requirements for an accurate robot movement and an acceptable response time for control commands and supervision. It is expected that the results of this paper can contribute with research groups about agricultural mobile robots and CAN-based distributed control technologies providing knowledge and enabling these implementation. 5. ACKNOWLEDGEMENTS The authors acknowledge the FAPESP - The State of São Paulo Research Foundation and the CNPq The National Council for Scientific and Technological Development for the support to this paper. 6. REFERENCES Aroca, R. V., Analysis of real time operating systems for robotics and automation applications (In Portuguese), Dissertation (Master), Escola de Engenharia de São Carlos, USP, São Carlos, Brazil, Åstrand, B. and Baerveldt, A., An Agricultural Mobile Robot with Vision- Based Perception for Mechanical Weed Control, Autonomous Robots, Vol.13, No.1, July, pp Auernhammer, H. and Speckmann, H., Dedicated Communication Systems and Standards for Agricultural Applications, Chapter 7, Section 7.1 Communication Issues and Internet Use, ASAE CIGR Handbook of Agricultural Engineering. Vol. 7, pp Auernhammer, H., Off-Road Automation Technology in European Agriculture - State of the Art and expected Trends. Proceedings of the ASAE International Conference of Automation Technology for Off-road Equipment, Kyoto, Japan., ASAE Publ. #701P1004, pp Baillieul, J. and Antsaklis, P.J., 2007, Control and Communication Challenges in Networked Real Time Systems, Proceedings of IEEE Technology of Networked Control Systems, Vol. 95, No. 1, pp Bak, T. and Jakobsen, H., Agricultural Robotic Platform with Four Wheel Steering for Weed Control, Biosystems Engineering, Vol. 87, Issue 2, February, pp Benneweis, R.K., Status of the ISO11783 Serial control and communications data network standard, ASAE International Meeting. ASAE paper Florida. Blackmore, B.S., Fountas, S., Tang, L. and Have, H., Design specifications for a small autonomous tractor with behavioral control, Journal of the International Commission of Agricultural Engineering (CIGR) VI. July Blackmore, B.S., Stout, W., Wang, M. and Runov, B., Robotic agriculture the future of agricultural mechanization?, European Conference on Precision Agriculture. 5 ed. J. Stafford, V. The Netherlands, Wageningen Academic Publishers. pp Blackmore, S. M.; Griepentrog, H. W., Autonomous Vehicles and Robotics, Chapter 7, Section 7.3 Mechatronics and Applications, ASAE CIGR Handbook of Agricultural Engineering. Vol. 6, pp Bosch, 2006, CAN Specification Version 2.0, Available in: < Access in: July, Darr, M. J.; Stombaugh, T. S. and Shearer, S., CAN based distributed control for autonomous vehicles, Transactions of the ASAE, Vol. 48, pp Johansson, K.H., Torngren, M. and Nielsen, L., 2005, Vehicle applications of controller area network, Handbook of Networked and Embedded Control Systems, Ed. Birkhäuser, 25p. Keicher, R. and Seufert, H., Automatic guidance for agricultural vehicles in Europe, Computers and Electronics in Agriculture. Vol.25, No.1-2, pp Moyne, J.R. and Tilbury, D.M., 2007, The Emergence of Industrial Control Networks for Manufacturing Control, Diagnostics, and Safety Data. IEEE Technology of Networked Control Systems, Vol. 95, No. 1, pp Nagasaka, Y.; Zhang, Q.; Grift, T. E.; Kanetani, Y.; Umeda, N. and Kokuryu, T., An autonomous field watching-dog robot for information collection in agricultural fields, ASAE Annual International Meeting, 2004, Ottawa, Ontario, Canada. ASAE paper No August. Pedersen, S.M.; Fountas, S.; Have, H. and Blackmore, B.S., Agricultural robots: an economic feasibility study, Precision Agriculture. Vol. 5, pp Reid, J.F.; Zhang, Q; Noguchi, N. and Dickson, M., Agricultural automatic guidance research in North America, Computers and Electronics in Agriculture. Vol.25, No.1-2, pp Sousa, R. V., CAN (Controller Area Network): an approach to control and automation in agricultural area (In Portuguese), Dissertation (Master), Escola de Engenharia de São Carlos, USP, São Carlos, Brazil, 94p. Suvinen, A. and Saarilahti, M., Measuring the mobility parameters of forwarders using GPS and CAN bus techniques, Journal of Terramechanics. Vol. 43, pp RESPONSIBILITY NOTICE The authors are the only responsible for the printed material included in this paper.